A necessary condition for the success of any machine learning model is that it achieves an accuracy that is superior to pre-existing methods. In the healthcare sphere, however, accuracy alone does not, nor should it, ensure that a model will gain clinical acceptance. So, what constitutes a good machine learning model for clinical applications? Unlike problems outside of the medical domain, poor performance for clinical models can have deleterious consequences for patients. In view of the fact that no model, in practice, has 100% accuracy, attempts to understand when a given model is likely to fail should form an important part of the evaluation of any machine learning model that will be used clinically. Models that provide physiologically motivated explanations for a given prediction are useful because they enable clinicians to leverage their understanding of the underlying physiology to gauge whether a given prediction is likely to be correct. In this talk I will describe novel approaches for building models that are motivated by our understanding of cardiovascular pathophysiology to produce predictions and associated explanations that are easily understood by the practicing clinician. It is our view that such models have a higher chance of being embraced, and used, by the clinical community.
Physiology-inspired machine learning models for predicting adverse cardiovascular outcomes
September 30, 2020
Dept. of Electrical Engineering and Computer Science, Institute for Medical Engineering and Science, Computer Science and Artificial Intelligence Laboratory, MIT; Massachusetts General Hospital